g.l
Well-known member
Haven't used that, but looks like it's also AVR based.
GH1 Stereoscopic Rig Collaboration
- Thread starter David Cole
- Start date
Haven't used that, but looks like it's also AVR based.
pixelpartner
New member
As a firmware coder for other micro processors and user of StereoDataMaker (6x TX1 = TX6), I also don't think it's 100% achievable to synch both (or more) GH1 cams perfectly for more than 10 Secs. On the other hand, a pair of TX1, running SDM and triggered by a short 5V pulse on USB, keep synch rather good. That's probably because it uses a rather outdated, but good per frame compression (Motion JPEG)
Have you thought about always feeding the GH13 shots thru a retimer application?
Theese can be used to do motion analysis and object recognition and massage any timing differences gracefully. The foundry has both - plugins and the compositing suite nuke - to achieve this. But FUSION should do the trick as well.
Have you thought about always feeding the GH13 shots thru a retimer application?
Theese can be used to do motion analysis and object recognition and massage any timing differences gracefully. The foundry has both - plugins and the compositing suite nuke - to achieve this. But FUSION should do the trick as well.
Last edited:
David Cole
Well-known member
As far as I know, there aren't any motion-flow analysis retimers that are stereo-aware. That process is non-deterministic - so the inevitable artifacts (e.g. tearing and occlusions) will introduce dissymmetry. We've use Nuke and Ocula for re-convergence and have had to deal with such induced artifacts.
I'm not aware of any commercially available tool for post that can fix intra-frame sync errors. They are baked in.
Synch on GH1s is almost perfect for still photos, at least according to some reports, like with TX1s. However, I cannot see why a par of TX1s would be much better than a pair of GH1s in VIDEO mode. Namely, I was involved in developing tight synchronisation for SDM so I know what I am talking about. By the way, compression has NOTHING to do with synch.
If they can create "more than perfect" intra-frame pictures, then I believe that such SW could solve the problem. However, I did not find any similar application yet.
AlanHunter
Member
External box for sync
External box for sync
Displaying the error could be done with an external box that monitors the composite outs. I'm sure they already exist, but I could theoretically build one (I'm a programmer and I've dabbled with microprocessors a bit).[/quote]
I've got a prototype external sync box built. I just got a GH1, haven't had time to see if it plays nice with the sync box. Will try it by the weekend. Details on the sync box at www.syncamajig.com
Glad to see here that feeding the GH1 power while the power switch is on does not seem to do any harm.
Alan
External box for sync
Displaying the error could be done with an external box that monitors the composite outs. I'm sure they already exist, but I could theoretically build one (I'm a programmer and I've dabbled with microprocessors a bit).[/quote]
I've got a prototype external sync box built. I just got a GH1, haven't had time to see if it plays nice with the sync box. Will try it by the weekend. Details on the sync box at www.syncamajig.com
Glad to see here that feeding the GH1 power while the power switch is on does not seem to do any harm.
Alan
David Cole
Well-known member
Let's go with the Arduino, then. Besides, PICBasic Pro is $249 just for the compiler. If anyone has a suggestion for the solid-state switches, that would be GREAT. We need a fast switch capable of handling 9.3V at 400Ma. Suppose we can just use a transistor?
Last edited:
g.l
Well-known member
The problem is not maintaining running sync a-la genlock. That would of course be the preferred solution, but the cameras will only drift apart very slowly. So as long as the capture start can be tightly synchronised, that would be fine, except for very long takes (> 10-20 mins).
I've never used SDM. Can you tell us about how this sync mechanism works for video? You mention the USB pulse, but how does the firmware manage to synchronise video capture to that? Is it able to adjust capture speed on the fly in software?
Crunchy, you dark horse! Great, so how did you do it for SDM?
I can't see it happening for stereo, it doesn't even work reliably in 2D.
Awesome Alan, look forward to your results. What cameras have you been testing with, and what sort of drift are you finding with them?
Hmm, they talk about their own programming language. They're also dev boards that are more expensive & with probably far more stuff than you need. Unless it has lots of advantages (it may), I recommend sticking with plain AVRs. I tend to use the Tiny 8bit / 6-8pin ones. They're really small, you can buy them cheap, and C(++) is my language. But whatever works for you David.
Edit: Free AVR GCC compiler.
David Cole
Well-known member
Hey... I have a little good news. It seems that the GH1s continue to output the vertical blanking pulse on the composite video output, even when they are recording (no video, however). If this is a useful measure of sync (testing now, but it takes a LONG time), we'll be able to monitor during recording as well.
Another oddity... the "trash button for 10 seconds" trick seems to stick. I've only done it once and the cameras are still outputting composite video after many power cycles. I wonder how you turn this off?
Another oddity... the "trash button for 10 seconds" trick seems to stick. I've only done it once and the cameras are still outputting composite video after many power cycles. I wonder how you turn this off?
g.l
Well-known member
Good find! I wonder why though?
g.l
Well-known member
I wanted to mention another feature the firmware hack could help with, as it just came up on the 3Dtv list - in-camera window placement.
Right now, stereo cameras are either placed with their lenses completely parallel, or with the lenses angled in slightly to converge at a particular distance (aka as 'toe-in').
In the first case, you end up loosing resolution by having to crop the views horizontally in post to set the stereo 'window' (ie. which parts of the image will be perceived at the screen plane of the display device). The toe-in method can avoid this (you're effectively setting the window during shooting), but it requires post-correction of the distortion introduced, which also looses resolution & quality.
You can avoid quality loss by shooting in parallel, but using shift-lenses to set the desired convergence/window.
A cheap & more reliable solution would to record a sub-rectangle off the sensor, and allow the user to move this horizontally in-camera to set the convergence - in effect you trade a little FOV for in-camera windowing. OR, capture a wider-than-needed rectangle (say 2400x1080), so the window can be set entirely in post without any quality loss.
Vitaliy, you've already experimented with capturing non-standard resolutions & sub-rectangles off the sensor - do you think either of these might be possible?
Right now, stereo cameras are either placed with their lenses completely parallel, or with the lenses angled in slightly to converge at a particular distance (aka as 'toe-in').
In the first case, you end up loosing resolution by having to crop the views horizontally in post to set the stereo 'window' (ie. which parts of the image will be perceived at the screen plane of the display device). The toe-in method can avoid this (you're effectively setting the window during shooting), but it requires post-correction of the distortion introduced, which also looses resolution & quality.
You can avoid quality loss by shooting in parallel, but using shift-lenses to set the desired convergence/window.
A cheap & more reliable solution would to record a sub-rectangle off the sensor, and allow the user to move this horizontally in-camera to set the convergence - in effect you trade a little FOV for in-camera windowing. OR, capture a wider-than-needed rectangle (say 2400x1080), so the window can be set entirely in post without any quality loss.
Vitaliy, you've already experimented with capturing non-standard resolutions & sub-rectangles off the sensor - do you think either of these might be possible?
David Cole
Well-known member
I'm failing to get a valid measurement of sync differences by looking at the vertical blank in the composite video signal on our DSO. I'm triggering on the negative slope of the blank from the LEFT camera - and can see the vertical blank from the RIGHT camera parade across the screen. I cannot, however, figure out how to measure the offset. This scope won't capture both channels and doesn't offer a chop display.
I can probably get my hands on a dual-input frequency counter... but it is not at my immediate disposal. Testing might have to move on to something else, for now.
A friend suggested the following part to switch the power and remote lines (ST Microelectronics STP12PF06 MOSFET). Looks like a winner to me.
I can probably get my hands on a dual-input frequency counter... but it is not at my immediate disposal. Testing might have to move on to something else, for now.
A friend suggested the following part to switch the power and remote lines (ST Microelectronics STP12PF06 MOSFET). Looks like a winner to me.
AlanHunter
Member
Syncamajig
Syncamajig
Awesome Alan, look forward to your results. What cameras have you been testing with, and what sort of drift are you finding with them?
I bought about ten Canon HV20's, and an HV30 and an HV40, and of the dozen I got two pair that are very well-matched. All of these Canon's start very consistently. I posted some details on my website. There are graphs (under "Photos") showing how each camcorder drifts versus a black-burst generator. Some pairs drift as little as 20 microseconds a minute. The oscillator frequency changes as the cameras warm up. I also bought a bunch of Samsung HMX10's, and got maybe one well-matched pair. These do not start very consistently. I also bought a few Casio EH-FH20's (for shooting high-speed video), and got one well-matched pair. On the Canon HV20's, I've been experimenting with starting them with matching cadence for 24PsF output. The only pair I've tested so far seems to start with matching cadence, but I've got a long list of things to try. I'll try some things with the GH1 on the weekend. Alan
Syncamajig
Awesome Alan, look forward to your results. What cameras have you been testing with, and what sort of drift are you finding with them?
I bought about ten Canon HV20's, and an HV30 and an HV40, and of the dozen I got two pair that are very well-matched. All of these Canon's start very consistently. I posted some details on my website. There are graphs (under "Photos") showing how each camcorder drifts versus a black-burst generator. Some pairs drift as little as 20 microseconds a minute. The oscillator frequency changes as the cameras warm up. I also bought a bunch of Samsung HMX10's, and got maybe one well-matched pair. These do not start very consistently. I also bought a few Casio EH-FH20's (for shooting high-speed video), and got one well-matched pair. On the Canon HV20's, I've been experimenting with starting them with matching cadence for 24PsF output. The only pair I've tested so far seems to start with matching cadence, but I've got a long list of things to try. I'll try some things with the GH1 on the weekend. Alan
zcream
Well-known member
Just use this circuit and build yourself a sync comparator box. It will tell you when the drift is too much.
http://www.ledametrix.com/syncshep/index.html
Really easy and Rob Crockett already put up details of the parts.
http://www.ledametrix.com/syncshep/index.html
Really easy and Rob Crockett already put up details of the parts.
David Cole
Well-known member
Excellent! Thanks!
Funny - I emailed Rob two days ago to see if the video blanking sync co-varies with sync value that the LANC Shepard reports. I believe that the LANC Shepard (and other stereo LANC controllers) use the packet timing to calculate sync offset. Haven't heard back yet.
bastienfrancois
New member
guys,
this is brilliant, and I got a few questions:
can you explain:
6. Dual trace oscilloscope: what is it for?
7. I suspect the break out cable so you can record sound while still being able to trigger, correct?
8. why brreak out cable for AV Out/Digital jack?
9. How is your twin shutter release working?
help much apreciated.
i am at the early stages of a similar experience, just aquiered a second "second hand" hackable gh1 and actually built a small beamsplitter (handheld was the key) for it, the first results (synched in post) where, well, not very conclusive, although with static shots, and slow movements, it looked already pretty cool. I have been looking at the lanc chepherd but it does not seem to work with the gh1 since panasonic does not work with the lanc protocol I have been told. Looking forward to read what has been your solution.
this is brilliant, and I got a few questions:
can you explain:
6. Dual trace oscilloscope: what is it for?
7. I suspect the break out cable so you can record sound while still being able to trigger, correct?
8. why brreak out cable for AV Out/Digital jack?
9. How is your twin shutter release working?
help much apreciated.
i am at the early stages of a similar experience, just aquiered a second "second hand" hackable gh1 and actually built a small beamsplitter (handheld was the key) for it, the first results (synched in post) where, well, not very conclusive, although with static shots, and slow movements, it looked already pretty cool. I have been looking at the lanc chepherd but it does not seem to work with the gh1 since panasonic does not work with the lanc protocol I have been told. Looking forward to read what has been your solution.
Last edited:
David Cole
Well-known member
The scope is useful for looking at the various signals coming from the GH1s to find an indication of the level of synchronization between them. This is R&D... once we figure out what the "indicator" is, the scope won't be needed. All bets are on the composite video out, at the moment.
Yes, and for measuring the camera's output on these lines for possible indication of sync.
For measuring sync - see above.
Hard to say at the moment. We're not sure that the trigger-releease timing is a factor in sync. The mechanical double-throw switch probably needs to be replaced with a more precise solid-state switch. Stay tuned.
David Cole
Well-known member
Rob Crocket says that the LANC Shepard and the Sync Shepard's video sync measurements match, however, the LANC Shepard is much more precise. Probably if you calibrate the proportional voltage output of his Sync Shepard, you'll get a very precise measure.
David Cole
Well-known member
I've got a Arduino test harness breadboarded out that will measure the voltage from Rob Crocket's Sync Shepherd circuit and latch 4 MOSFET switches independently (LEFT & RIGHT power, LEFT & RIGHT shutter). Waiting on a few components from Digi-Key to complete.
I'm going to control it all from the PC at first, with a terminal program. We can move on to an easier (and cheaper) to duplicate controller if this bears fruit.
I'm going to control it all from the PC at first, with a terminal program. We can move on to an easier (and cheaper) to duplicate controller if this bears fruit.
Last edited:
AlanHunter
Member
Syncamajig and GH1
Syncamajig and GH1
I'll try some things with the GH1 on the weekend. Alan[/quote]
Crunchy's suggestion of holding down the "trash" button for 10 seconds works. I get composite sync out, and the LCD viewfinder still displays. I'm wondering whether this impacts the video quality. The Panasonic engineers presumably had a reason for shutting off composite output while recording. Also, I'm wondering who found out that holding down the trash button for 10 seconds would do this! What in the world made them try it? Was it in a service manual? I tried to test my GH1 for consistent startup times, but all my other camcorders run on 7.4 volts, and the GH1 does not start unless it is getting over 9 volts. (oddly, the Panasonic battery that comes with it says right on it that it is a 7.4 volt battery). Instead I get a message on the LCD VF like "cannot use this battery". The Syncamajig uses one external battery to drive both camcorders, so I'm not likely to know much more until I get a 2nd GH1 and a high-capacity external 9.3 volt battery. That will be weeks from now. Or I might buy a cheap little 9V NTSC surveillance camera just for this test. I did hook up 3 regular 9V batteries in parallel, and ran the GH1 for over an hour watching its drift against a blackburst generator. It warms up and becomes very stable after about 30 minutes (that is, the oscillator frequency stops rising). I also used 3 fresh 9V Energizer batteries to start the GH1 repeatedly using the Syncamajig, and that worked fine for about 18 restarts, by which time it had already drained the 9V batteries enough that it said "cannot use this battery". (I find that Energizers really do have more energy than most other 9V alkalines - 3 new industrial-strength Ray-o-vac batteries only allowed 5 restarts.) The point of this was just to get more confidence that turning the camera on by feeding it power while the power switch was on was not harmful. I could have used the AC adapter, except that I had already cut the DC coupler cable and soldered on a barrel connector sized for the Syncamajig. I will only use battery power for other tests and actual use, if any, as AC adapters introduce noise that introduces randomness in startup times. Alan
Syncamajig and GH1
I'll try some things with the GH1 on the weekend. Alan[/quote]
Crunchy's suggestion of holding down the "trash" button for 10 seconds works. I get composite sync out, and the LCD viewfinder still displays. I'm wondering whether this impacts the video quality. The Panasonic engineers presumably had a reason for shutting off composite output while recording. Also, I'm wondering who found out that holding down the trash button for 10 seconds would do this! What in the world made them try it? Was it in a service manual? I tried to test my GH1 for consistent startup times, but all my other camcorders run on 7.4 volts, and the GH1 does not start unless it is getting over 9 volts. (oddly, the Panasonic battery that comes with it says right on it that it is a 7.4 volt battery). Instead I get a message on the LCD VF like "cannot use this battery". The Syncamajig uses one external battery to drive both camcorders, so I'm not likely to know much more until I get a 2nd GH1 and a high-capacity external 9.3 volt battery. That will be weeks from now. Or I might buy a cheap little 9V NTSC surveillance camera just for this test. I did hook up 3 regular 9V batteries in parallel, and ran the GH1 for over an hour watching its drift against a blackburst generator. It warms up and becomes very stable after about 30 minutes (that is, the oscillator frequency stops rising). I also used 3 fresh 9V Energizer batteries to start the GH1 repeatedly using the Syncamajig, and that worked fine for about 18 restarts, by which time it had already drained the 9V batteries enough that it said "cannot use this battery". (I find that Energizers really do have more energy than most other 9V alkalines - 3 new industrial-strength Ray-o-vac batteries only allowed 5 restarts.) The point of this was just to get more confidence that turning the camera on by feeding it power while the power switch was on was not harmful. I could have used the AC adapter, except that I had already cut the DC coupler cable and soldered on a barrel connector sized for the Syncamajig. I will only use battery power for other tests and actual use, if any, as AC adapters introduce noise that introduces randomness in startup times. Alan